Wang Xinyu, Mi Yuanyuan, Zhang Zhaoyang, Chen Yang, Hu Gang, Li Haihong
School of Science, Beijing University of Posts and Telecommunications, Beijing 100876, China.
Center for Neurointelligence, School of Medicine, Chongqing University, Chongqing 400044, China and AI Research Center, Peng Cheng Laboratory, Shenzhen 518005, China.
Phys Rev E. 2022 Jul;106(1-1):014302. doi: 10.1103/PhysRevE.106.014302.
Quantitative research of interdisciplinary fields, including biological and social systems, has attracted great attention in recent years. Complex networks are popular and important tools for the investigations. Explosively increasing data are created by practical networks, from which useful information about dynamic networks can be extracted. From data to network structure, i.e., network reconstruction, is a crucial task. There are many difficulties in fulfilling network reconstruction, including data shortage (existence of hidden nodes) and time delay for signal propagation between adjacent nodes. In this paper a deep network reconstruction method is proposed, which can work in the conditions that even only two nodes (say A and B) are perceptible and all other network nodes are hidden. With a well-designed stochastic driving on node A, this method can reconstruct multiple interaction paths from A to B based on measured data. The distance, effective intensity, and transmission time delay of each path can be inferred accurately.
包括生物和社会系统在内的跨学科领域的定量研究近年来备受关注。复杂网络是进行此类研究的常用且重要的工具。实际网络产生的数据呈爆炸式增长，从中可以提取有关动态网络的有用信息。从数据到网络结构，即网络重构，是一项关键任务。在实现网络重构方面存在许多困难，包括数据短缺（存在隐藏节点）以及相邻节点之间信号传播的时间延迟。本文提出了一种深度网络重构方法，该方法即使在只有两个节点（例如A和B）可感知而所有其他网络节点均隐藏的情况下也能工作。通过在节点A上精心设计的随机驱动，此方法可以基于测量数据重构从A到B的多条交互路径。每条路径的距离、有效强度和传输时间延迟都可以准确推断出来。
